Hacker News front pageAug 31, 2026, 19:17 UTC+8

Agent memory as a file format beats complex pipelines and knowledge graphs

Agent Memory as a File Format

Cal Paterson proposes 'memoryfield': a zip of Markdown files plus an optional SQLite vector index for agent memory. Agents write short prose pages directly, then use semantic search to jump to all relevant pages in parallel—at most two tool calls. He argues this is faster and more reliable than knowledge-graph traversal, which adds 2–3 seconds per hop and often misses relevant info. The post doesn't include benchmark comparisons. Soft page limit is ~2,000 tokens.

Why it matters: A concrete, opinionated take on agent memory with a spec you could actually try — zip + Markdown + SQLite. Score stays at 72 because it's a personal blog proposal without an open-source implementation or community validation yet; it's a design argument, not a shipped artifact.

Read the original ↗Export Markdown

Take

I clicked this because it reframes agent memory as a data format problem, not a pipeline problem. Cal Paterson's pitch: let the agent write short Markdown pages directly, bundle them in a zip, and optionally add a SQLite vector index. Retrieval is a single semantic search—at most two tool calls—instead of walking a knowledge graph hop by hop.

His critique of graph traversal is the useful bit. Each hop adds 2–3 seconds, and the agent has to guess relevance from link text alone, which often misses things. I buy that argument—serial tool calls for page-hopping are slow and brittle in practice.

But I'd discount a few things. The post has zero benchmarks, just qualitative reasoning. The ~2,000-token soft limit per page is a real constraint for long-document use cases, and he acknowledges it. The whole approach also assumes the agent writes good memory pages upfront; if it misses key details at write time, semantic search won't save you later.

This reads more like a lightweight memory scheme for personal agents than a replacement for enterprise RAG. If you're building a small tool that needs to remember user preferences, this is way lighter than spinning up pgvector and Neo4j.

Source: Hacker News front page · calpaterson.com
Published: Aug 31, 2026, 19:17 UTC+8 · Aug 31, 2026